0.1.8 Beta5 gate: Publish the three-route maintained-Provider Responses candidate #351

Parent

#248

Release purpose

Publish 0.1.8-beta.5 after all generic Responses capabilities from Beta2 through Beta4 have been independently regression-tested against the three Ollama Cloud Provider/model configurations maintained by CodexHub:

• 0.1.8 Beta5: Run the full maintained GLM-5.2 Provider regression #275 — GLM-5.2
• 0.1.8 Beta5: Run the full maintained K2.7 Code Provider regression #276 — K2.7 Code
• 0.1.8 Beta5: Run the full maintained DeepSeek V4 Flash 0731 Responses regression #393 — DeepSeek V4 Flash 0731 (deepseek-v4-flash:0731)

Beta5 adds no planned generic capability. It validates the complete generic implementation against maintained real-world Responses routes.

Acceptance criteria

• The exact Beta4 candidate lineage is preserved.
• 0.1.8 Beta5: Run the full maintained GLM-5.2 Provider regression #275, 0.1.8 Beta5: Run the full maintained K2.7 Code Provider regression #276, and 0.1.8 Beta5: Run the full maintained DeepSeek V4 Flash 0731 Responses regression #393 are closed with independent real-CLI evidence covering all shipped Beta2–Beta4 Responses workflows.
• All three maintained routes exercise the same generic type/protocol implementation, including the complete Collaboration V2 lifecycle.
• Production code contains no GLM/K2.7/DeepSeek model-name dispatch or runtime allowlist.
• A Provider-specific documented endpoint/auth/protocol fact may live in maintained configuration, but tool conversion remains generic.
• Exact Provider/model identity remains unchanged and no request reaches a second Provider for tool execution.
• Failures distinguish upstream behavior, maintained configuration, transport availability, and generic CodexHub defects.
• Beta1 through Beta4 regressions remain green.
• Exact candidate SHA review, CLI/manual verification, and local verification/release gates succeed.
• Only after those gates pass is 0.1.8-beta.5 tagged and published as a manual opt-in prerelease.

Out of scope

• Runtime model eligibility or advertising CodexHub as supporting only these models.
• Adding a new generic workflow during the release gate.
• Chat compatibility and full Chat Collaboration V2, which are Beta6.
• Cross-Provider tool proxying.

Blocked by

• 0.1.8 Beta4 gate: Publish the runtime-derived generic Collaboration V2 candidate #284
• 0.1.8 Beta5: Run the full maintained GLM-5.2 Provider regression #275
• 0.1.8 Beta5: Run the full maintained K2.7 Code Provider regression #276
• 0.1.8 Beta5: Run the full maintained DeepSeek V4 Flash 0731 Responses regression #393